A multiply-accumulate, or MAC, is an operation that multiplies two numbers and adds the result to a running total. It is the core step of the matrix math at the heart of AI models, where large numbers of such operations are combined to produce each result.

Chip designers build hardware around MAC units, and performance ratings are often derived from how many a chip can perform per second. Because each MAC combines a multiplication and an addition, some vendors count it as two operations when quoting TOPS, and others count it as one.

That difference in counting is one reason peak figures from different vendors may not be directly comparable. A MAC rating also says nothing about whether the rest of the system can keep the units supplied with data, which depends on memory and software.

When reading a spec, check how operations are counted and at what precision. Treat the figure as a statement about theoretical capacity and look for real-world tests for how the chip behaves on actual models.